presented by Hageun Park in Asia Water Week 2013, Manila 11-15 March 2013

1. A Comparative Study of Social Behavior inA Comparative Study of Social Behavior in
Irrigated and Rain‐fed Areas: The Case of Bohol
Irrigation Scheme, the Philippinesgat o Sc e e, t e pp es
Hogeun Park, Takuji W. Tsusaka, and Valerien O. Pede

2. Introduction
 To investigate the connection between management of
l ( it ) i i ti d f ’ i l b h icanal (gravity) irrigation and farmers’ social behavior
(1) Measures social behavior through behavioral game
experimentsexperiments
(2) Estimates the effects of irrigation, neighborhood,
as well as individual characteristics.
 Combination of 1) behavioral game experiments and
2) hierarchical linear modeling
 The availability of irrigation water in the village does The availability of irrigation water in the village does
not only improve agricultural productivity but also
enhances social relationship among farmers

3. Overview of Bohol Irrigation Project
• The Bohol Irrigation System, located in the northeastern part g y , p
of Bohol Island about 50 km from the provincial capital city of
Tagbilaran, began operation in May 2008
• JICA did feasibility study in 1985
• San Miguel Ubay and Trinidad• San Miguel, Ubay, and Trinidad
• Gravity irrigation system by
Bayongan dam
• Service Area 3,295ha
• 17.5km of Main Canal

4. Structure of Dataset
 IRRI conducted
• Agricultural and Socioeconomic Data (X)
4 crop seasons from 2009 to 2010
4-season Average
Behavioral Game Results (Y)• Behavioral Game Results (Y)
Sep. 2011
290 randomly selected farmers290 randomly selected farmers
Irrigated (N = 144) & Rain-fed (N = 146)

5. Theoretical Framework
• Behavioral game experiments are designed so as tog p g
quantify participants’ social behavior under strategic
situations (Gintis 2003).
• Employing dictator game and ultimatum game, which are
developed to explore altruistic and retaliating behaviors,developed to explore altruistic and retaliating behaviors,
respectively

6. Behavioral Game Experiments
 Dictator Game
• This game is intended to elicit participants’ fairness,
generosity, or altruism (Hoffman et al., 1996).
??100 PHP is equivalent to 2.46 (USD) by Bloomberg currency data, as of 31 January 2013. The Philippines’ GDP per
capita is $2,370 (2011) as per World Bank data. Given these exchange rate and GDP per capita, 100 PHP is considered
sufficient to ensure incentive compatibility for the experiment purpose

7. Behavioral Game Experiments
 Ultimatum Game
• This game is interpreted as an indicator of the receiver’s
retaliating behavior or unwillingness to tolerate the level
of distribution (Herbert et al 2003)of distribution (Herbert et al., 2003).
?x x?x x100 PHP is equivalent to 2.46 (USD) by Bloomberg currency data, as of 31 January 2013. The Philippines’ GDP per
capita is $2,370 (2011) as per World Bank data. Given these exchange rate and GDP per capita, 100 PHP is considered
sufficient to ensure incentive compatibility for the experiment purpose

8. Results for Behavioral Game Experiments
Type of Anonymous
(1) Irrigated (2) Rain‐fed (3) t‐test for
Type of Anonymous
Partner
Sample
(N=131)
Sample
(N=114)
mean difference
|(1)‐(2)|
Dictator Game
Someone in Sender’s
Purok
33.97 27.81 6.16**
(20.59) (19.04) [0.015]
Someone in Sender’s
B
32.06 27.11 4.96*
(21 58) (18 28) [0 053]Barangay (21.58) (18.28) [0.053]
Ultimatum Game
Someone in Sender’s
Purok
24.43 34.83 10.40***
(15 15) (19 61) [0 000]Purok (15.15) (19.61) [0.000]
Someone in Sender’s
Barangay
25.12 34.47 9.36***
(16.47) (21.29) [0.000]

9. HLM (Hierarchical Linear Modeling)
• While ANOVA and OLS analyses are commonly used iny y
quantitative assessments, care must be taken when the
data are nested (Raudenbush and Byrk 1993).
• “Frog-Pond” Theory; Robinson(1950) the problem of
contextual effects
Reference: J. Kyle Roberts., An introduction to HLM with R
http://faculty.smu.edu/kyler/training/AERA_overheads.pdf

10. HLM (Hierarchical Linear Modeling)
• Our data set covers randomly selected 238 rice farmersy
who reside in 3 municipalities and 18 barangays
• Altruistic and retaliating behaviors arise from social• Altruistic and retaliating behaviors arise from social
atmosphere; we try to differentiate individual effects
from barangay effects
• Employing HLM to account for the barangay-level
characteristics that are expected to affect individualp
level social behaviors

11. Descriptive Statistics
Level 1 (Household Level)
i blVariable N Mean SD Min Max
Age 238 51.38 12.06 14 87
Schooling Years 238 6.33 3.02 0 14
Asset Holding (Log PhP) 238 10.61 1.09 6.21 13.31g ( g )
Household Size 238 5.93 2.32 1 12.5
Parcel Size (ha) 238 1.45 1.02 0.12 8.12
Level 2 (Barangay Level)
blVariable N Mean SD Min Max
Irrigation Dummy 18 0.61 0.5 0 1
Age 18 51.3 4.5 43.56 61
Schooling Years 18 6.37 0.93 4.46 8Schooling Years 8 6.37 0.93 4.46 8
Asset Holding (Log PhP) 18 10.57 0.52 9.44 11.53
Household Size 18 5.99 1.1 4.65 8.76
Parcel Size (ha) 18 1.31 0.46 0.58 2.19

12. Estimates for Intercept‐only Model
Random
Coefficient
St. Dev.
Variance
Component
d.f. 2 p‐value ICC
Dictator Game
Intercept 1, u0 5.830 33.989 17 38.817 0.002 0.085
Level‐1, r 19.079 364.008
Ultimatum Game
INTRCPT1 u 6 668 44 463 17 49 456 <0 001 0 120INTRCPT1, u0 6.668 44.463 17 49.456 <0.001 0.120
Level‐1, r 17.725 314.163
12

13. Estimates for level‐1 Equations
[Level-1 Equation]
β β ( ) β (S ) β ( ) β ( S ) βYij = β0j + β1j (Ageij) + β2j (Schooling Yearsij) + β3j (Assetij) + β4j (Household Sizeij) + β5j
(Parcel Sizeij) + rij
[Level-2 Equation]
β β β β β ββ0j = γ00 + u0j, β1j = γ10 + u1j, β2j = γ20 + u2j, β3j = γ30 + u3j, β4j = γ40 + u4j, β5j = γ50 + u5j
Game
Type
β0
(Intercept 1)
Age
Schooling
Years
Asset
Household
Size
Parcel
SizeType (Intercept 1) Years Size Size
Dictator 28.789*** -0.268*** 0.109 -0.658 0.143 0.375
2Ultimatum 28.117*** -0.067 -0.578* -1.984* -0.427 0.797
*** p < 0.01, * p < 0.10
1
2
3

14. Estimates for level‐2 Equations
Yij = γ00 + γ01 (Irrigation Dummyj) + γ02 (Agej) + γ03 (Schooling Yearj) + γ04 (Assetj) +
γ05 (Household Sizej) + γ06 (Parcel Sizej) + γ10 (Ageij) + γ20 (Schooling Yearij) +
γ30 (Assetij) + γ40 (Household Sizeij) + γ50 (Parcel Sizeij) + u0j + u1j (Ageij) + u2j
(Schooling yearij) + u3j (Assetij) + u4j (Household Sizeij) + u5j (Parcel Sizeij)+ rij
Game
Type
γ00
(Intercept 2)
Irrigation
Dummy
Age
Schooling
Years
Asset
House
hold
Size
Parcel
Size
Dictator 23.387*** 9.053* 0.166 -0.259 4.348* -0.724 6.087
Ultimatum 39 092*** -14 012*** -0 697** -1 124 -8 585*** 0 885 -4 964Ultimatum 39.092 -14.012 -0.697 -1.124 -8.585 0.885 -4.964
*** p < 0.01, ** p <0.05, * p < 0.10 1 23

15. Concluding Remarks
• The result is highly suggestive of the significant socialg y gg g
effects of canal irrigation schemes.
• The positive effect on altruism and the negative effect onThe positive effect on altruism and the negative effect on
retaliation indicate that the type of social interactions
promoted by the necessity for collective irrigation
management leads to inducing the accumulation ofmanagement leads to inducing the accumulation of
“good” social behavior among farmers.
O l t lid ti th i i ti ff t i t id• One clue to validating the irrigation effect is to consider
the existence of TSAs (turnout service associations)
in the irrigated communitiesg

16. Concluding Remarks
• TSA
- private canal construction
- purchasing machinery
providing micro credit- providing micro credit
Compared with the rain-fed, irrigated farmers are
d i i d diexposed to more opportunities to meet and discuss
public arrangements with their neighbors
• Dual role:
to boost the rural economy through increased production,
and to accumulate social capital among farmersand to accumulate social capital among farmers.

17. Anecdotal Information
• Cultivated cassava before irrigation project
• Cultivating Hybrid Rice twice a year
• Three children
- Crop science
V t i- Veterinary
- Agronomy
• Promoting children back to village for
( Inday Salaum )
• Promoting children back to village for
agriculture
• Several neighbors’ children already back
to village for their career( Inday Salaum ) to village for their career
• Irrigation and modern agricultural technology cang g gy
prevent brain drain from rural areas.

18. Limitation
• Our behavioral game experiments were conducted ing p
2011 which was after the construction of irrigation. This
survey structure prevents us from formulating a
difference in difference estimator that ensures a moredifference-in-difference estimator that ensures a more
proper impact assessment.